The numbers are stark. Core Scientific, one of North America’s largest Bitcoin miners, paid Block, Inc. (formerly Square) $41.9 million to walk away from a contract for 15 exahash worth of next-generation 3-nanometer mining chips. Not a renegotiation. Not a delay. A clean termination with a nine-figure penalty. The message is unmistakable: Block’s Proto chip programme was so uncompetitive that Core preferred burning $41.9 million in cash to deploying the hardware.
This is not a story about a single failed commercial deal. It is the autopsy of a strategic miscalculation by Jack Dorsey, and a clear signal that the Bitcoin mining industry is undergoing a structural pivot toward AI infrastructure—one that leaves traditional miners scrambling for relevance. As an on-chain detective who has spent years auditing consensus mechanisms and tokenomics, I have seen this pattern before. The ledger does not forgive. And here, the ledger shows a clear exit signal.
Context: The Hype Behind Block’s Mining Ambitions
In 2023, Block announced its venture into Bitcoin mining silicon—a 3nm ASIC chip developed by its Proto team. The narrative was compelling: Jack Dorsey, Bitcoin maximalist CEO of a $40 billion payments company, would bring his engineering talent to break the duopoly of Bitmain and MicroBT. Core Scientific, a major miner with tens of thousands of machines, was announced as the launch customer. The deal promised up to 15 EH/s of compute power, enough to rival the fleet sizes of Marathon or Riot.
But the crypto industry runs on delivery, not promises. Verification precedes trust. By early 2025, the first units were deployed. Then the silence. Core Scientific’s Q1 2025 earnings revealed the $41.9 million charge for contract termination. The official reason was a "strategic pivot" toward AI computing. But reading between the lines, the chip’s real-world performance—specifically its energy efficiency (J/TH) and stability—likely fell short of expectations. In a sector where every joule matters, a 10% efficiency gap can destroy margins.
Core: Systematic Teardown of the Failure
Let’s examine the evidence piece by piece.
First, the financial loss. Core Scientific paid $41.9 million to cancel. That represents the total prepayment for the chips plus a penalty. For a company that emerged from bankruptcy in early 2024 with restructuring debt, that is a material sum. Yet they chose to take the hit rather than accept the hardware. Why? Because deploying suboptimal miners would have incurred greater opportunity cost: lost revenue, higher electricity bills, and competitive disadvantage against miners using Bitmain’s S21 or MicroBT’s M60 series, which boast superior J/TH.
Second, the timing. The contract was signed in 2023. By late 2024, Bitmain had already shipped high-volume 3nm units with proven field performance. Block’s chip, if it existed at all in volume, was late to market. In the semiconductor world, being late is death. Core Scientific had a window to deploy and generate revenue; Block missed it.
Third, the strategic pivot. Core Scientific simultaneously announced a 15-year, $14.8 billion contract with AMD to provide AI data center hosting. The company is converting its existing mining facilities into high-performance computing (HPC) centers. This is not a side project; it is a transformation. The $41.9 million write-off is the price of clearing the balance sheet for this new direction. It is a bet that AI rents will exceed Bitcoin mining profits for the foreseeable future.
I have seen similar dynamics before. In 2020, during DeFi Summer, I audited Curve Finance’s stableswap invariant and discovered exploitable rounding errors that would trigger under high volatility. Everyone was yelling about yield farming; I was yelling about formal verification. The market eventually cared—after the hacks. Here, the market is ignoring that a high-profile hardware project just failed its first real customer. Follow the coins, not the claims.
The Unravelling of Block’s Crypto Empire
Block’s mining chip failure is not an isolated incident. It sits within a pattern of failed or abandoned crypto initiatives under Jack Dorsey: Tidal (music platform) written down to near zero, TBD (decentralized identity) abandoned, Bitkey (self-custody wallet) struggling, even the "Bitchat" messaging app launched and killed. The company also paid $200+ million in fines to the CFPB for Cash App fraud issues. Over the past five years, Block’s stock has declined 68%. Code is law. Logic is lethal. And the logic here points to a CEO whose crypto bets have consistently failed to deliver.
From a technical perspective, designing competitive ASIC miners requires hundreds of millions in R&D and years of iteration. The incumbents—Bitmain with 70% market share, MicroBT with 20%—have been refining their architectures for a decade. Block tried to skip the line with a single custom chip. Arrogance, not innovation, best describes the approach.
The Contrarian Angle: What the Bulls Got Right
Admittedly, there is an argument that Block’s chip technology might have been sound, and Core Scientific’s exit is purely about AI opportunity, not chip quality. The AI bubble is real; companies are paying exorbitant prices for GPU clusters. Core could have deployed both Bitcoin miners and AI racks, but chose to focus. That is a business decision, not a technology indictment.
Furthermore, Block’s Proto team may still find other customers—maybe smaller miners willing to take a chance on a new supplier if price is low enough. After all, Bitmain also had early teething problems with its 7nm chips.
But the weight of evidence leans against that narrative. No other major miner stepped in to take Core’s allocation. Block has not announced any alternative buyer. The $41.9 million penalty suggests Block was unable to remarket the chips at a price that recouped the prepayment. In a market where second-hand S19s trade at $15/TH, Block’s 3nm chips, if sold at a discount, would need to be drastically cheaper than Bitmain’s flagships. That destroys the unit economics of the entire programme. Verification precedes trust—and the market has verified that Block’s chips are not trusted.
Takeaway: Accountability Call
The Bitcoin mining industry is at a crossroads. Core Scientific’s choice to absorb a $41.9 million loss and shift to AI is a loud declaration that Bitcoin-only mining is no longer the optimal use of capital and energy. For miners who fail to adapt, the coming cycle could be brutal. For Block, the crypto hardware chapter may be closed. Investors should watch for writedowns in Block’s next 10-K as the Proto business line is likely impaired or sold.
The ledger does not forgive. It records every contract, every failure, and every exit. This one is written in red ink to the tune of $41.9 million. Follow the coins, not the claims. The coins, in this case, never arrived.